Low-cost BLE based Indoor Localization using RSSI Fingerprinting and Machine Learning

Abstract

With the IoT industry undergoing a boom, Location Based Services (LBSs) are playing an essential role in constructing smart cities. Since LBSs are in the process of being ubiquitous, it is essential to find a low-cost and low energy solution for localization. Bluetooth Low Energy (BLE) technology for indoor localization is a smart choice for giving economical solutions to the industry with additional advantages like trouble-free connection to other gadgets. We propose an improved RSSI (Received Signal Strength Indicator) based fingerprinting technique in which data is first augmented and then classified using Machine Learning algorithms. This indoor localization technique facilitates us to recognize the XY-position of the user node (UN) or tag, which receives signals from the anchor nodes (ANs). The fluctuations in the RSSI values were large, because they were affected by multipath propagation and other factors in the indoor positioning environment. The fingerprinting data was augmented to overcome these drawbacks, to decrease the computational cost, to guarantee the precision of the framework by increasing the accuracy and also to be equipped for Machine Learning calculation. The augmentation strategy was actualized by utilizing accessible RSSI esteems at one location, wherein Random Forest gave the highest test accuracy of 96% surpassing all existing methods.